This is just a super quick demo for GPT-J
(included in the Huggingface transformers
package)
if anyone wants to try it out.
Not suitable for anything other then experimentation and demo.
Pay attention that this repo does not contain a lot of things including:
- Optimizations
- Error handlings
# Build
docker-compose build
# Run
docker-compose up -d
# Then wait until each service is fully started (e.g.: GPT-J model needs to be loaded)
Then you can access the following:
- Swagger UI:
localhost:8008/docs
- Streamlit App:
localhost:8050